معرفی
Konstantin Sering is a researcher at the University of Tübingen within the Faculty of Philosophy, specifically working at the Seminar for Linguistics as part of the Quantitative Linguistics group. His academic journey bridges physics and psychology with computational linguistics, culminating in his 2023 PhD thesis on the PAULE (Predictive Articulatory speech synthesis Utilizing Lexical Embeddings) model. He maintains active collaborations with the Cognitive Modeling Group led by Martin V. Butz and the Quantitative Linguistics Group led by Harald Baayen.
Sering's research interests center on the computational modeling of human speech production, with particular focus on the articulatory aspects of language. He approaches spoken language from the perspective of computer-aided modeling and simulation, extending the VocalTractLab model to produce spontaneous human speech. His work emphasizes the information-bearing nature of language over its meaning-bearing aspects, investigating how language signals contextual states and conditions. He combines artificial neural networks with predictive forward approaches to model vocal tract control parameters across multiple target spaces, including meaning and acoustic representations.
Analysis of Sering's publication record reveals a consistent trajectory in articulatory speech synthesis and quantitative linguistics. His work demonstrates increasing sophistication in modeling the human speech apparatus, with recent publications focusing on improving the PAULE model through speech/non-speech classification, somatosensory feedback integration, and benchmarking methodologies. His research spans both theoretical computational linguistics and practical applications, with strong connections to cognitive science and psycholinguistics. The publications show interdisciplinary collaboration across linguistics, computer science, and cognitive psychology.
Sering actively contributes to the academic community through teaching and software development. He has taught courses including Introduction to Psycholinguistics, Physics of Human Speech Production, and seminars on Predictive Articulatory Speech Synthesis. As software author and maintainer, he has developed several important tools for the field: Articubench (an articulatory speech synthesis benchmark), PAULE (his namesake speech synthesis model), create_vtl_corpus, and contributions to pyndl and other packages. His technical expertise spans Python, R, and specialized instrumentation for speech analysis.
Based in room 188 at Keplerstr. 2 in Tübingen, Sering maintains an active research program with regular publications and software releases since 2013. His work environment benefits from the resources of the University of Tübingen's linguistics infrastructure and collaborative opportunities across disciplines.
